patrick-kidger / patrick-kidger/diffrax

BacksolveAdjoint() fails to compute gradients

Open
#596 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
2.1k
Forks
189
Avg merge
3d 18h
Merged PRs (30d)
1

Description

Hi,
I have been encountering a new issue with diffrax version 0.6.2, that is when I try to use BacksolveAdjoint() to compute gradients, the gradients value exploded to infinity even on reversible ODE problems. When I downgraded diffrax to version 0.6.1 or 0.6.0, I don't encounter this problem anymore. I will provide a minimal example later in this issue thread.

Thanks!

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Reproduce the reported gradient explosion with BacksolveAdjoint() on a reversible ODE, once the minimal example is available. Compare behavior in diffrax 0.6.2 with 0.6.1 and 0.6.0; the issue is done when gradients remain finite in the affected case.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.